v0.1.0-alpha.2: /events + KPI avanzate + E2E + idle-timeout + fix ESM

Delta da alpha.1

WS /events endpoint

• WebSocket /events integrato nel daemon (ku + health + ack)
• Token bucket backpressure (10 msg/s, burst 20)
• Idle timeout WS (5 min) + ping/pong per client inattivi
• ACK immediato incrementa metrics.net.acked

Metrics avanzate

• Prometheus exposition: /metrics?format=prom
• Nuove metriche: sgn_net_delivered/acked/dedup_ratio, sgn_ws_clients, sgn_events_drop
• Istogrammi con p50/p95 per latenza HTTP
• CI smoke: artifact prom.txt automatico

E2E tests robusti

• events-minimal-e2e: publish → WS ku → ack → metrics
• events-slow-client-e2e: client lento, health continua, no loss KU
• metrics-*-e2e: contatori/quantili/Prometheus
• Anti-race: waitForHealth() + path assoluti

VSCode extension

• Subscription automatica a /events
• Toast "KU ricevuta" con bottoni Open/Verify
• Settings: sgn.eventsPath (default /events)
• README con troubleshooting e esempi

Fix & hardening

• Fix ESM: require() → statSync in sqlite-real-storage
• Log daemon_listen per readiness
• Cleanup processi child nei test

Pilot ready

Quick start:

npm run daemon:start
wscat -c ws://localhost:8787/events  # vedi health + ku
curl -s http://localhost:8787/metrics?format=prom

KPI da monitorare:

• Delivery: sgn_net_acked / sgn_net_delivered ≈ 1.0
• Latenza: sgn_http_publish_p50 < 200ms, p95 < 1s
• Stabilità: sgn_events_drop ≈ 0

Pilot setup:

• trust.json: mode warn per 24h → poi enforce
• VSCode extension: toast real-time su KU
• CI: KU-receipt on test fail

Vedi README_PILOT.md per setup completo.